Unfortunately I can locate none at the moment. My main reason for believing such is my own experience when I replicated a test with a 10 segment LED bar graph, with the segments driven in 10% duty increments, and to my eye the gradations were closer to logarithmic (I had printouts of a linear and logarithmic scale, both in 10 gradations to match to). The phenomenon may be due to some other factor but I have seen it in practice.

A photograph of the experiment showed a linear scale, but this was not as it appears to the eye, probably due to some process in the eye or the brain in processing the image.